RESEARCH ARTICLE Open Access

Full-length transcriptome sequencingreveals the low-temperature-tolerancemechanism of Medicago falcata rootsGuowen Cui1†, Hua Chai1,2†, Hang Yin1, Mei Yang1, Guofu Hu1, Mingying Guo3, Rugeletu Yi3 and Pan Zhang1*

Abstract

Background: Low temperature is one of the main environmental factors that limits crop growth, development,and production. Medicago falcata is an important leguminous herb that is widely distributed worldwide. M. falcatais related to alfalfa but is more tolerant to low temperature than alfalfa. Understanding the low temperaturetolerance mechanism of M. falcata is important for the genetic improvement of alfalfa.

Results: In this study, we explored the transcriptomic changes in the roots of low-temperature-treated M.falcata plants by combining SMRT sequencing and NGS technologies. A total of 115,153 nonredundantsequences were obtained, and 8849 AS events, 73,149 SSRs, and 4189 lncRNAs were predicted. A total of 111,587 genes from SMRT sequencing were annotated, and 11,369 DEGs involved in plant hormone signaltransduction, protein processing in endoplasmic reticulum, carbon metabolism, glycolysis/gluconeogenesis,starch and sucrose metabolism, and endocytosis pathways were identified. We characterized 1538 TF genesinto 45 TF gene families, and the most abundant TF family was the WRKY family, followed by the ERF, MYB,bHLH and NAC families. A total of 134 genes, including 101 whose expression was upregulated and 33whose expression was downregulated, were differentially coexpressed at all five temperature points. PB40804,PB75011, PB110405 and PB108808 were found to play crucial roles in the tolerance of M. falcata to lowtemperature. WGCNA revealed that the MEbrown module was significantly correlated with low-temperaturestress in M. falcata. Electrolyte leakage was correlated with most genetic modules and verified that electrolyteleakage can be used as a direct stress marker in physiological assays to indicate cell membrane damage fromlow-temperature stress. The consistency between the qRT-PCR results and RNA-seq analyses confirmed thevalidity of the RNA-seq data and the analysis of the regulatory mechanism of low-temperature stress on thebasis of the transcriptome.

Conclusions: The full-length transcripts generated in this study provide a full characterization of thetranscriptome of M. falcata and may be useful for mining new low-temperature stress-related genes specificto M. falcata. These new findings could facilitate the understanding of the low-temperature-tolerancemechanism of M. falcata.

Keywords: Medicago falcata, Low-temperature stress, SMRT, RNA-seq, WGCNA

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: zhangpgrass@126.com†Guowen Cui and Hua Chai contributed equally to this work.1Department of Grassland Science, College of Animal Science andTechnology, Northeast Agricultural University, Harbin 150030, ChinaFull list of author information is available at the end of the article

Cui et al. BMC Plant Biology (2019) 19:575 https://doi.org/10.1186/s12870-019-2192-1

BackgroundLow temperature is one of the main environmental factorsthat limit plant growth, development and geographicaldistribution [1]. Low-temperature stress, which consists ofchilling stress (< 10 °C) and freezing stress (< 0 °C), can re-duce crop productivity to some extent [2]. The effects oflow temperature on plants depend on developmental stageand exposure time. Under low-temperature stress, youngtissues and organs are more severely damaged than oldtissues and organs, and the reproductive stage is moresensitive to low temperature than the vegetative stage is[3]. Exposure to low temperature causes physiological andmolecular changes in plants that lead to metabolic disor-ders: these changes include an inhibition of photosyntheticactivity; a reduction in water uptake; an increase in oxida-tive stress via increased accumulation of reactive oxygenspecies (ROS); an increase in intracellular pH and osmoticpressure; and functional abnormalities in chloroplasts,mitochondria and other organelles [4–6]. Furthermore,low temperature temporarily inhibits sucrose synthesis,and the rearrangement of membranes changes the stabilityand mobility of proteins and a shift in redox homeostasis,which decrease enzyme activities and alters both meta-bolic homeostasis and gene transcription [7, 8].Plants have developed many mechanisms and pathways

that enable them to minimize the negative effects of lowtemperature and to successfully grow and reproduce [9].Osmolytes induced by low-temperature stress, includingproline, soluble sugars, and cold-induced stress proteins(dehydrins and LEA proteins), can improve the osmoticpotential of cells, protect the stability of biological mem-branes, alleviate oxidative damage limitations, and evenact as signals to regulate the expression of stress-relatedgenes [10, 11]. The overexpression of SlNAM1, which en-codes a typical NAC, improves low-temperature tolerancein transgenic tobacco by improving osmolytes and redu-cing the H2O2 and superoxide anion radical (O2

.-) con-tents under low temperature, which contribute toalleviating the oxidative damage of the cell membraneafter low-temperature stress [12]. Low temperature in-duces the production of Ca2+, which can be sensed by cor-responding receptors, among which lipid Ca2+ channelsmay be the primary cryogenic signal receptors, and canthen activate calcium response protein kinase (CPKs,CIPKs, and CRLK1) and MAPK cascade responses, whichin turn regulate the expression of cold-responsive (COR)genes [13, 14]. The overexpression of COLD1 (jap) signifi-cantly increases chilling tolerance; COLD1 interacts withthe G-protein alpha subunit to activate Ca2+ channels tosense low temperature and to accelerate G-proteinGTPase activity [15]. Low-temperature stress rapidly in-duces the expression of many transcription factors (TFs),including CBF AP2-domain proteins, which then activatethe expression of numerous downstream COR genes [16–

18]. The expression of the CBF gene is controlled by up-stream TFs, such as the bHLH TF ICE1. ICE1 is subjectedto sumoylation and polyubiquitylation and subsequentproteasomal degradation mediated by the SUMO E3 ligaseSIZ1 and the ubiquitin E3 ligase HOS1, respectively [13,16].Single-molecule long-read (SMRT) sequencing, which

was developed by PacBio Biosciences RSII, represents athird-generation sequencing platform. Owing to its longreads, this platform is widely used in genome sequen-cing, and by generating full-length or long sequences, ithas eliminated many restrictions associated with sequen-cing [19–22]. Next-generation sequencing (NGS) tech-nology (RNA sequencing [RNA-seq]) can provideexpression profiles of both coding or noncoding RNAsand can generate digital data of gene expression, enab-ling rapid and cost-effective genomic and transcriptomicstudies for most major crop species, including rice [23],wheat [24], and grape [25]. The approach by which NGSand SMRT sequencing are combined has been appliedfrequently to generate comprehensive information at thetranscriptional level, allowing the identification of keyfunctional and regulatory genes involved in abiotic stressresistance [26–28]. The changes in gene expression inresponse to low temperature stress revealed by tran-scriptome analyses including TFs, protein kinases, smallnon-coding RNAs and enzymes in metabolic pathways,which regulated the expression of downstream genes,target genes, and protected cells from low-temperaturedamage [29, 30].Medicago falcata L., an economically and ecologically

important legume herb with an expanse from northernMediterranean regions to northern Russia, is closely re-lated to alfalfa but exhibits better tolerance to lowtemperature than alfalfa [31–33]. Understanding themechanism of low-temperature tolerance in M. falcata isimportant for the genetic improvement of alfalfa, which isthe most important forage leguminous species because ofits high biomass productivity, optimal nutritive profile andadequate persistence [34]. Although the low-temperaturetolerance of M. falcata is a popular research topic at themorphological level to the physiological, biochemical andmolecular biology levels [31, 35–37], few studies have in-vestigated the low-temperature tolerance of M. falcata atthe transcriptome level. In this paper, we combined SMRTand NGS to generate expression profiles of M. falcataroots under low-temperature stress. In total, 115,153 non-redundant sequences were obtained fromM. falcata roots,and 11,369 differentially expressed genes (DEGs) wereidentified, including 134 genes that were differentiallycoexpressed at all five temperature points. These findingsprovide a global characterization of gene transcription andfacilitate the understanding of the low-temperature-tolerance mechanisms of M. falcata.

Cui et al. BMC Plant Biology (2019) 19:575 Page 2 of 16

ResultsPhysiological responses of M. falcata under low-temperature stressWe measured the electrolyte leakage, malondialdehyde(MDA) content, superoxide dismutase (SOD), catalase(CAT) and peroxidase (POD) activities as well as thesuperoxide anion radical (O2

.-), soluble protein, reducedglutathione (GSH), proline and soluble sugar contents toinvestigate the physiological changes in M. falcata roots

exposed to low-temperature stress for 2 h (Fig. 1). Underlow-temperature stress, the electrolyte leakage increasedgradually with decreasing temperature (Fig. 1a). The great-est electrolyte leakage was observed at − 15 °C, which was4.18 times greater than that under the control conditions.The MDA content decreased gradually, peaked at − 10 °C(relative to that of the control) and slightly decreased at −15 °C (Fig. 1b). No obvious difference in SOD activity wasobserved, and the activity increased after low-temperature

Fig. 1 Determination of physiological indices of the roots of M. falcata plants under low-temperature stress. a, Electrolyte leakage. b, MDAcontent. c, SOD activity. d, CAT activity. e, POD activity. f, O2

.- content. g, Soluble protein content. h, GSH content. i, Proline content. j, Solublesugar content. The data are shown as the means ± SDs of four independent experiments. The different letters represent statistically significantdifferences as determined by one-way ANOVA (p < 0.05, Duncan’s multiple range test)

Cui et al. BMC Plant Biology (2019) 19:575 Page 3 of 16

treatment and decreased from 4 °C to − 15 °C (Fig. 1c).The changes in CAT and POD activities were com-pletely different under different temperatures. CAT ac-tivity first increased, whereas the POD activity firstdecreased (Fig. 1d-e). The content of O2

.- increased sig-nificantly after low-temperature treatment and peakedat − 10 °C (Fig. 1f). The content of soluble protein de-creased first after low-temperature treatment but thenincreased after 4 °C, after which it decreased signifi-cantly after 0 °C (Fig. 1g). The GSH content increasedsignificantly after low-temperature treatment andpeaked at − 5 °C (Fig. 1h). The proline content de-creased slightly in the low-temperature environment(Fig. 1i), and the greatest soluble sugar content wasmeasured at − 10 °C (Fig. 1j).

M. falcata transcriptome sequencingTo identify and characterize the transcriptome of M. fal-cata roots under low-temperature stress, we measured theroots under room temperature (CK), 4 °C, 0 °C, − 5 °C, −10 °C and − 15 °C for 2 h by combining the PacBio SMRTand NGS technologies for whole-transcriptome profiling.In total, 125.48 Gb of clean data were obtained by RNA-seq, yielding 418,495,643 reads (with a GC content of42.61% and a QC 30 of 87.55%) (Additional file 1: TableS1). A total of eight SMRT cells were used for the three li-braries at three size ranges, 1–2 kb, 2–3 kb, and 3–6 kb,yielding 19.27 Gb of clean data. A total of 1,202,336 poly-merase reads were obtained, and the polymerase reads con-taining fragments that were less than 50 bp in length andwith a sequence accuracy lower than 0.75 were subse-quently filtered and removed. A total of 8,428,385 subreadswere then obtained by filtering the remaining sequencesfrom the linkers, the linker sequences and the subreads

with fragments less than 50 bp in length (Table 1). A totalof 552,818 reads of inserts (ROIs, of which 270,750 werefull-length nonchimeric reads (FLNCs), and 223,319 werenon-full-length reads, were extracted from the original se-quence (Table 1). The full-length sequences were clusteredvia the RS_IsoSeq module of SMRT Analysis software. Atotal of 131,118 consensus isoforms were obtained; 99,490high-quality isoforms were obtained via non-full-length se-quence alignment, and 31,628 low-quality isoforms wereobtained and corrected via RNA-seq data. Any redundancywithin the high-quality and corrected low-quality transcriptsequences of each sample was eliminated by CD-HIT soft-ware, and 115,153 nonredundant sequences were obtained.On the basis of the nonredundant sequences of each sam-ple, we predicted a total of 8849 alternative splicing (AS)events via the IsoSeq_AS_de_novo script (Additional file 2:Table S2); 73,149 simple sequence repeats (SSRs) via theMIcroSAtellite identification tool (Additional file 3: TableS3); and 4189 long noncoding RNAs (lncRNAs) with thecoding potential calculator (CPC), coding–non-codingindex (CNCI), coding potential assessment tool (CPAT),and Protein family (Pfam) database information(Additional file 4: Figure S1).

Annotation and expression of transcripts under low-temperature stressTo acquire the most comprehensive annotation informa-tion, all full-length transcripts from SMRT were alignedwith NCBI nonredundant protein (Nr), SwissProt, GeneOntology (GO), Clusters of Orthologous Groups (COG),Eukaryotic Ortholog Groups (KOG), Pfam, and KyotoEncyclopedia of Genes and Genomes (KEGG) databaseinformation via BLAST software (version 2.2.26), and atotal of 111,587 genes from SMRT were annotated, of

Table 1 Statistical results of the SMRT sequencing data

cDNA size 1–2 kb 2–3 kb > 3 kb All

SMRT cells 3 3 2 8

Polymerase reads 450,876 450,876 300,584 1,202,336

Postfilter number of subreads 4,761,465 2,481,575 1,185,345 8,428,385

Reads of insert 240,441 189,779 122,598 552,818

Number of 5′ reads 122,516 120,763 84,001 327,280

Number of 3′ reads 139,768 126,512 87,086 353,366

Number of poly-A reads 133,709 124,866 86,112 344,687

Number of filtered short reads 42,363 11,932 2853 57,148

Number of non-full-length reads 100,142 75,634 47,543 223,319

Number of full-length reads 97,936 102,213 72,202 272,351

Number of full-length nonchimeric reads 97,156 101,959 71,635 270,750

Average full-length nonchimeric read length 1309 2282 3536 2264

Full-length percentage (FL%) 40.73% 53.86% 58.89% 49.27%

Artificial concatemers (%) 0.80% 0.25% 0.79% 0.59%

Cui et al. BMC Plant Biology (2019) 19:575 Page 4 of 16

which the length of 7155 genes was > = 300 bp and <1000 bp and of which the length of 104,432 genes was> = 1000 bp (Additional file 5: Table S4). Among theseannotated sequences, 111,384 sequences had significantmatches in the Nr database, 89,117 sequences had sig-nificant matches in the Pfam database, 82,433 sequenceshad matches in the SwissProt database, and 3385 had ef-fective matches in the GO database. On the basis of thehomology among sequences of different species, 97,831(87.87%) sequences were found against M. truncatula,and 3463 (3.11%) sequences had clandestine hits withCicer arietinum, followed by M. sativa (1662, 1.49%),Glycine max (1011, 0.91%), Rhizoctonia solani (528,0.47%), Fusarium oxysporum (388, 0.35%), Glycine soja(369, 0.33%), Phaseolus vulgaris (349, 0.31%), Vitis vinif-era (266,0.24%), and M. falcata (181, 0.16%). Only 5283(4.75%) annotated sequences were similar to those ofother plant species (Additional file 6: Figure S2).To evaluate gene expression levels in response to low-

temperature stress, we mapped all the clean data back tothe assembled transcriptome, and the read count for eachgene was obtained from the mapping results via RNA-seqby expectation maximization (RSEM) software(Additional file 7: Table S5). The mapped read count foreach gene was then converted to the expected number offragments per kilobase of transcript per million mappedreads (FPKM) (Additional file 8: Table S6). FPKM valuescan eliminate the effects of transcript length and sequencingdifferences on computational expression. A boxplot diagramof the FPKM values indicated that gene expression levelswere not evenly distributed in the different experimental en-vironments (Fig. 2a). Pearson correlation coefficients weresubsequently used to evaluate the correlations of each bio-logical sample, with r2 values close to 1 indicating a strongcorrelation between two replicate samples (Fig. 2b). After-ward, all the sequences were then used for further DEGanalysis after abnormal samples were excluded.

Analysis of DEGs in response to low-temperature stressIn total, 11,369 DEGs whose expression was up- or down-regulated between samples (fold change≥2 and false discov-ery rate (FDR) < 0.01) at any pair of temperature points wereidentified by comparing gene expression levels under low-temperature stress (Additional file 9: Table S7). Clusteringpatterns of the DEGs of plants under low-temperature stresswere determined by hierarchical cluster analysis of all theDEGs (Fig. 3a). All DEGs exhibiting the same or similar ex-pression levels were clustered, and it was determined that aset of genes was quickly expressed during the early stage oflow-temperature stress (4 °C) and that other genes wereexpressed under freezing temperature (− 10 °C). The 11,369DEGs identified were grouped into six subclusters by K-means coexpression cluster analysis (Fig. 3b). The expres-sion level of genes in subcluster 1 (1271 genes) began to

increase after the temperatures decreased to − 5 °C andreached a maximum at − 10 °C; the expression levelsthen decreased rapidly. KEGG analysis of the genes insubcluster 1 revealed that most were involved instarch and sucrose metabolism, plant-pathogen inter-actions, and galactose metabolism. The expression ofgenes in subcluster 2 (2226 genes) increased signifi-cantly after low-temperature treatment, after whichthe level first decreased slowly and then rapidly ateach temperature point below 4 °C. Genes in this sub-cluster functioned mostly in the circadian rhythm,plant-pathogen interactions, and plant hormone signaltransduction. The expression of the genes in bothsubcluster 3 (1460 genes) and subcluster 4 (2695genes) was upregulated after low-temperature treat-ment and downregulated after 4 °C treatment. Thedifference was that the decrease in the expression ofgenes in subcluster 3 fluctuated, while that in sub-cluster 4 was continuous. Genes involved in starchand sucrose metabolism, protein processing in endo-plasmic reticulum, fatty acid degradation, and tyrosinemetabolism were enriched in subcluster 3, and genesin subcluster 4 were enriched in plant-pathogen inter-actions, starch and sucrose metabolism and plant hor-mone signal transduction. The expression of genes insubcluster 5 (1613 genes) was weakly downregulatedafter low-temperature treatment, strongly upregulatedfrom 4 °C to 0 °C, and then upregulated again under− 15 °C treatment. The genes in this subcluster functionedmostly in plant-pathogen interactions, phenylalanine me-tabolism, and the plant hormone signal transduction path-way. The expression of genes in subcluster 6 (2104 genes)was downregulated at all times, and most of these genesfunctioned in phenylpropanoid biosynthesis, circadianrhythm, and ubiquitin-mediated proteolysis.

Identification of putative TFsTFs play an important role in cell function and develop-ment and directly regulate gene expression via interac-tions with themselves and other proteins to participatein plant stress regulation, including low-temperature-related processes. In this study, 1538 TF genes were dif-ferentially expressed between different temperaturepoints and were classified into 45 TF gene families ac-cording to information in the PlantTFDB (http://planttfdb.cbi.pku.edu.cn/) (Additional file 10: Table S8).The most abundant TF family was the WRKY (186genes) family, followed by the ERF (165 genes), MYB(143 genes), bHLH (131 genes) and NAC (111 genes)families. The cluster analysis of TF gene expression indi-cated that the expression of some of these TF genes wasextensively upregulated in response to low temperature(Additional file 11: Figure S3), such as those encodingbZIP, AP2/ERF, MYB, C2H2 and WRKY and TFs.

Cui et al. BMC Plant Biology (2019) 19:575 Page 5 of 16

Fig. 2 Comparison of gene expression levels under low-temperature stress. a, Boxplot showing the distribution of the FPKM values of eachsample under low-temperature stress. The X-axis in the boxplot shows the ID of each sample. The Y-axis represents the log10(FPKM). b, Heat mapof the Pearson correlation coefficient of each sample under low-temperature stress

Cui et al. BMC Plant Biology (2019) 19:575 Page 6 of 16

Comparison of coexpressed genes between low-temperature-treated samples and control samplesUnder low-temperature stress, a total of 8683 DEGswere identified by a comparison of each temperaturepoint to the control environment. Interestingly, therewere many more upregulated DEGs (5876 genes) thandownregulated DEGs (2807 genes), and 134 genes weredifferentially coexpressed at all five temperature points(Additional file 12: Table S9). As shown in Fig. 4a, therewere 101 upregulated genes and 33 downregulated genesacross all five comparisons. These 134 coexpressed geneswere assigned to the biological process, cellular compo-nent and molecular function GO categories (Fig. 4b). Inthe biological process category, “metabolic process”, “cel-lular process”, “single-organism process” and “biologicalregulation” were the most enriched terms. In the cellularcomponent category, “cell”, “cellular component” and“organelle” were the most enriched, and in the molecularfunction category, “binding” was the most enriched term,followed by “catalytic activity”. COG functional classifi-cation of the 134 coexpressed genes showed that mostof the genes were enriched in “general function

prediction only”, “transcription”, “replication, recombin-ation and repair”, “signal transduction mechanisms” and“carbohydrate transport and metabolism” (Fig. 4c).KEGG enrichment revealed that most of the coexpressedgenes were enriched in the “circadian rhythm - plant”,“cysteine and methionine metabolism”, “arginine andproline metabolism”, “phenylpropanoid biosynthesis”,“galactose metabolism”, “plant-pathogen interaction”and “biosynthesis of amino acids” pathways (Fig. 4d).

Weighted gene coexpression network analysis (WGCNA)of DEGs in response to low-temperature stressWGCNA was performed to better understand whichgenes within these complex signaling networks were themost connected hubs. The number of genes in the mod-ule were clustered according to their expression levels,and those genes with a high clustering degree were allo-cated to the same models. The 8683 DEGs identified bycomparing low-temperature-treated samples and controlsamples were clustered on the basis of topological over-lap, and then the gene modules were obtained from adynamic tree cut. Finally, 12 gene modules were

Fig. 3 Clustering analysis of the DEGs. a, Hierarchical clustering of the 11,369 DEGs on the basis of the averaged log2(FPKM+ 1) values of allgenes in each cluster. b, The six subclusters of the 11,369 DEGs were clustered. The number of genes in each subcluster is shown at the top ofthe subcluster. The blue line shows the average values of the relative expression levels in each subcluster, and the gray lines represent therelative gene expression levels of each gene in each subcluster

Cui et al. BMC Plant Biology (2019) 19:575 Page 7 of 16

identified after merging modules whose expression pat-terns were similar (Fig. 5a). The magenta modules con-tained the most genes (1092), and the violet modulescontained the fewest (70) (Additional file 13: Table S10).The gray module was not a true module but a place tocategorize the remaining genes that were not well corre-lated with any one of the significant colored modules. Themodule eigengene-based connectivity (kME) value wascalculated for each gene to every module, and 449 geneswere found to act as a hub in more than one module.All 12 genetic modules with a module characteristic

value of p < 0.05 were used to identify the modulesthat were highly correlated with samples and physio-logical indicators (Fig. 5b). The control samples werehighly correlated with the MEblue module. The sam-ples under 4 °C treatment were highly correlated withthe MEmagenta module, and the samples under 0 °Ctreatment were highly correlated with the MElightc-yan module. The samples under − 5 °C treatment werehighly correlated with the MElightyellow module, thesamples under − 10 °C treatment were highly corre-lated with the MEbrown module, and the samplesunder − 15 °C treatment were highly correlated withthe MEgray60 module. The MEbrown module wasfound to be significantly correlated with physiologicalindicators and may play a key role in the low-temperature tolerance of M. falcata. The COG

classification results showed that the MEbrown mod-ule genes were involved in 23 major categories, in-cluding “general function prediction only”, “signaltransduction mechanisms”, “transcription” and “replica-tion, recombination and repair” (Additional file 14: FigureS4A). The GO analysis results showed that the MEbrownmodule genes were involved mainly in 13 biological pro-cesses, including “protein phosphorylation”, “regulation oftranscription, DNA-templated”, and “oxidation-reductionprocess”. These genes were distributed to eight cellularcomponent terms, including “integral component ofmembrane”, “plasmodesma”, “chloroplast stroma” and“nucleus”, and were associated with 14 molecular func-tions, which included ATP binding, protein serine/threo-nine kinase activity, and cation binding (Additional file 14:Figure S4B). KEGG enrichment analysis showed that mostof the genes in the MEbrown module were enriched in“starch and sucrose metabolism”, “plant-pathogen inter-action”, “circadian rhythm - plant”, “protein processing inendoplasmic reticulum”, “galactose metabolism” and“plant hormone signal transduction” (Additional file 14:Figure S4C).

Confirmation of RNA-seq sequencing data by qRT-PCRanalysisThe DEGs associated with low-temperature stress wereselected for qRT-PCR assays to confirm the SMRT

Fig. 4 Summary of coexpressed genes between low-temperature-treated samples and control samples. a, Venn diagram of DEGs identified by acomparison of each temperature point to the control environment. b, GO classification of 134 coexpressed genes. c, COG functional classificationof 134 coexpressed genes. d, KEGG enrichment analysis of 134 coexpressed genes

Cui et al. BMC Plant Biology (2019) 19:575 Page 8 of 16

Fig. 5 WGCNA of the genetic modules related to each sample and physiological indicators. a, Cluster dendrogram. b, Module–trait relationships

Cui et al. BMC Plant Biology (2019) 19:575 Page 9 of 16

sequencing data. Twenty genes were selected randomlyfrom 134 DEGs coexpressed at all five temperaturepoints. We found that the fold-changes in expressioncalculated via the sequencing data did not exactly matchthe expression values detected by qRT-PCR analysis, butthe expression profiles were essentially consistent for all20 genes (Additional file 15: Figure S5). These analysesconfirmed the reliability of the gene expression valuesgenerated from the SMRT sequencing data.

DiscussionThe approach of combining NGS and SMRT sequencinghas become increasingly popular for studying plant re-sponses to adverse environmental conditions to providehigh-quality and increasingly complete assemblies at thetranscriptome level [26–28]. SMRT sequencing can gen-erate full-length or long sequences, and the high errorrate can be overcome via short and high-accuracy NGSreads [19, 22]. In this study, we combined the SMRTand RNA-seq methods to analyze the transcriptome as-sembly of roots of M. falcata plants under low-temperature stress and to identify key functional andregulatory genes involved in low-temperature tolerance.We ultimately obtained 115,153 nonredundant se-quences, and the average ROI was long enough to repre-sent the full-length transcripts (Table 1). We alsopredicted a total of 8849 AS events, 73,149 SSRs, and4189 lncRNAs.

Changes in physiological indicatorsOur results indicated the complexity of physiologicalchanges within M. falcata plants in response to low-temperature stress. Under low-temperature stress, theroots displayed relatively more extensive changes inmembranes, antioxidants and osmolytes. An increasein the contents of MDA, proline, and soluble sugarsas well as electrolyte leakage has been demonstratedin cold-treated wheat and wild tomato [25, 38]. Dur-ing low-temperature treatment, M. falcata accumu-lates relatively high amounts of sucrose and prolineand exhibits high sucrose phosphate synthase (SPS)and sucrose synthase activity [39]. In this paper, anincrease in MDA, proline and soluble sugar contentsas well as electrolyte leakage was observed in theroots of low-temperature-treated M. falcata plants,suggesting that osmolytes might protect plant cellmembranes, increase membrane stabilization, and bal-ance osmotic pressure during low-temperature-induced dehydration of M. falcata; additionally, elec-trolyte leakage can be used as a direct stress markerto reflect cell membrane damage by low-temperaturestress (Fig. 1). Low-temperature stress induces the ac-tivities of cell apoptosis factors, and it has beenproven that injury to M. falcata under low

temperature is related to proteins in cells [40]. Theenzymatic antioxidant system is a protective mechan-ism used to eliminate or reduce ROS and increase aplant’s ability to tolerate low-temperature stress [41,42]. Thus, the changes in CAT, POD, and GSH mayplay a key role in the detoxification of ROS inducedby low temperature in M. falcata roots.

Gene expression in the roots of M. falcata in response tolow-temperature stressIn this study, a total of 11,369 DEGs were identified asresponsive to low-temperature stress at all temperaturepoints, of which the expression of 68.8% was inducedand that of 31.2% was repressed under low-temperaturestress. All DEGs were grouped into six subclusters (Fig.3b), after which KEGG pathway enrichment analysis wasconducted. In the organismal systems category, the mostenriched pathway was “plant-pathogen interaction”, indi-cating a basic plant immunological response in the rootsof M. falcata plants under low-temperature stress [43].In the category of environmental information processing,most DEGs were involved in the plant hormone signaltransduction pathway. Our results demonstrated thecritical role of phytohormones in plants in response toexternal and internal cues to regulate growth and devel-opment [36]. In the genetic information processing cat-egory, the most enriched pathway was “proteinprocessing in endoplasmic reticulum”. Under low-temperature stress, the membrane protein synthesis rateand membrane protein number increased in cold-adapted alfalfa [44]. Our data showed the facilitation andmonitoring of proper folding by chaperone interactionsand the formation of assemblies into multimeric proteinsin the endoplasmic reticulum. In the metabolism cat-egory, most DEGs were involved in “carbon metabol-ism”, followed by “glycolysis/gluconeogenesis” and“starch and sucrose metabolism”, suggesting that carbonand energy supplies were very important for the adapta-tion of M. falcata to low temperatures. In the cellularprocess category, most DEGs functioned in endocytosis.Endocytosis regulates the entry of membrane proteins,lipids, and extracellular molecules into the cell under ad-verse environmental conditions and plays a key role inalleviating ROS [45, 46].

Genes encoding TFs in response to low-temperaturestressTFs play an important role in cell function and develop-ment and directly regulate gene expression via interac-tions with themselves and proteins to participate inplant stress regulatory processes, including low-temperature-related processes. Many TFs, includingbHLH, bZIP, MYB, C2H2, ERF, NAC and WRKY types,that confer low-temperature tolerance to plants have

Cui et al. BMC Plant Biology (2019) 19:575 Page 10 of 16

been identified via transcriptomic approaches [47]. Inthis study, 1538 TF genes were differentially expressedbetween different temperature points and were classifiedinto 45 TF gene families. The most abundant TF familywas the WRKY family, followed by the ERF, MYB,bHLH and NAC families, and the dynamic changes ingene expression associated with these TFs may revealtheir vital function in M. falcata low-temperature toler-ance. Our results are consistent with those of previousreports on TFs involved in plant responses to low-temperature stress [31, 36] and suggest that members ofthe WRKY family play a critical role in M. falcata low-temperature tolerance.

Identification of genes responsible for the response tolow temperaturePlants have developed many mechanisms and path-ways that enable them to minimize the negative ef-fects of low temperature. Global analysis of stress-responsive genes facilitates the understanding of plantresponses to low-temperature stress. In this paper, atotal of 8683 DEGs were identified by a comparisonof each temperature point to the control environ-ment, and 134 genes, including 101 whose expressionwas upregulated and 33 whose expression was down-regulated, were determined to be differentially coex-pressed at all five temperature points (Fig. 4).However, only 7 genes were successfully annotatedvia GO enrichment analysis, and two genes (PB40804and PB75011) were enriched in all three categories.The expression of both of these genes was upregu-lated under low-temperature stress. The product ofPB40804 functions as a phenylalanine ammonia-lyase,which acts as a smart switch directly controlling theaccumulation of calycosin and calycosin-7-O-beta-D-glucoside in Astragalus membranaceus plants duringlow-temperature treatment [48]. The PB75011 proteinfunctions as a decarboxylase and is involved in cellwall/membrane/envelope biogenesis. Ornithine de-carboxylase and arginine decarboxylase control thesynthesis of polyamines in plants. The response ofArabidopsis thaliana to low-temperature stress em-phasizes the involvement of transcriptional regulationin arginine decarboxylase gene expression [49]. Onthe basis of the COG annotation, we identified 18enriched genes, including 15 whose expression wasupregulated and 3 whose expression was downregu-lated. PB40804 was also annotated as “amino acidtransport and metabolism” according to the COGanalysis. The abundance of amino acid transporters iscorrelated with a multitude of fundamental roles inplant growth and development, and low-temperaturestress could decrease the amino acid concentrationsand alter their composition [50, 51]. “Circadian

rhythm - plant” was the most enriched pathway ac-cording to the KEGG results, and we found thatPB110405, the gene associated with the most GOterms in the biological process category, was enrichedin this pathway. PB110405 was annotated with 9 GOterms in the biological process category, including“regulation of transcription, DNA-templated”,“temperature compensation of the circadian clock”,“response to hydrogen peroxide”, “starch metabolicprocess” and “response to cold”. The association withthe “circadian rhythm-plant” pathway suggested thatthe internal temperature of M. falcata was substan-tially influenced by low temperature. The expressionof PB108808, a putative ortholog of a MYB-relatedLHY TF gene in Arabidopsis, is a gene that is in-duced by low temperature and that indicates the pres-ence of interplay between the circadian rhythm andthe response to low temperature in M. falcata. Thenext pathway was “arginine and proline metabolism”;arginine and proline metabolism is one of the centralpathways for the biosynthesis of the amino acids ar-ginine and proline [27]. Proline accumulation is awell-known means of alleviating abiotic stress inplants [52]. Combined with the changes in prolinecontents under low-temperature stress, our results in-dicated that osmotic regulatory substances, protectiveprotein molecules in M. falcata, play important rolesin the response to low-temperature stress, and thecomposition of aromatic compounds may changeunder low-temperature stress.

Identification of genetic modules corresponding to low-temperature stressWGCNA is a systematic biological method that canbe applied to the study of biological processes withmultiple sources [53]. It has been proven thatWGCNA can be an efficient data mining method,specifically for screening genes related to traits andfor conducting modular classification to determinecoexpression modules with high biological significance[54]. In this paper, 8683 DEGs were identified by acomparison of low-temperature-treated samples andcontrol samples and were clustered into 12 genemodules after the modules with similar expressionpatterns were merged together (Fig. 5). It was foundthat the MEbrown module was significantly correlatedwith low-temperature stress in M. falcata. GO enrich-ment analysis of the MEbrown module revealed thatregulatory pathways with biological significance couldbe obtained in this module. For example, the GO an-notation terms “cold-response pathways”, “response tostress” and “intracellular signal transduction” wereenriched. COG classification revealed that the MEb-rown module was enriched in many DEGs associated

Cui et al. BMC Plant Biology (2019) 19:575 Page 11 of 16

with general function prediction only and the signaltransduction mechanism, and KEGG pathway enrich-ment analysis revealed that there were many DEGsinvolved in starch and sucrose metabolism. We alsofound that electrolyte leakage was correlated withmore genetic modules than the other physiological in-dicators were, which corroborated our findings fromphysiological assays in that electrolyte leakage can beused as a direct stress marker of cell membrane dam-age from low-temperature stress.

ConclusionsOverall, by combining SMRT and NGS technologies, weexplored the transcriptomic changes in the roots of low-temperature-treated M. falcata plants. A total of 115,153nonredundant sequences were obtained, and 8849 ASevents, 73,149 SSRs and 4189 lncRNAs were predicted. Atotal of 111,587 genes from SMRT sequencing were anno-tated, and 11,369 DEGs were identified that are involvedin plant hormone signal transduction, protein processingin the endoplasmic reticulum, carbon metabolism, gly-colysis/gluconeogenesis, starch and sucrose metabolism,and endocytosis pathways. We characterized 1538 TFgenes into 45 TF gene families, and the most abundant TFfamily was the WRKY family, followed by the ERF, MYB,bHLH and NAC families. A total of 134 genes were differ-entially coexpressed at all five temperature points, includ-ing 101 genes whose expression was upregulated and 33genes whose expression was downregulated. PB40804,PB75011, PB110405 and PB108808 were found to playcrucial roles in the tolerance of M. falcata to lowtemperature. The WGCNA results showed that the MEb-rown module was significantly correlated with low-temperature stress in M. falcata. Moreover, electrolyteleakage was correlated with most genetic modules andverified that electrolyte leakage can be used as a directstress marker of cell membrane damage from low-temperature stress in physiological assays. These findingsprovide a complete characterization of gene transcriptionand facilitate the understanding of the mechanisms of tol-erance to low temperature in M. falcata.

MethodsPlant cultivation and low-temperature treatmentSeeds of M. falcata L. cv. Hulunbuir were collected fromthe Hulunbuir Grassland with the permission of theHulunbuir Grassland Station in the Inner MongoliaProvince of China. The Hulunbuir Grassland Station inthe Inner Mongolia Province of China undertook theformal identification of the samples and provided detailsof specimen deposited. It is a cultivated variety approvedby the National Grass Variety Approval Committee ofChina (Accession number: 269). Collection of the seedsneeds the permission of the Hulunbuir Grassland Station

in the Inner Mongolia Province of China. The experi-mental research on M. falcata complies with Chineseand international guidelines. Seeds of M. falcata weredisinfected with 5% sodium hypochlorite solution for 5min and then washed in distilled water. The seeds werethen germinated on wet filter paper in culture dishes ina growth chamber at 25 °C in the dark. Five-day-oldseedlings were transplanted into plastic pots that werefilled with a mixture of vermiculite:perlite:peat (1:1:1) inthe greenhouse, whose average temperature was 25 °Cand 20 °C and whose relative humidity was 55 and 70%during the day and night, respectively. All the seedlingswere watered with 1/2-strength Hoagland [55] nutrientsolution every two days. Ninety days after transplantation,uniform seedlings were transported to a growth chamberfor the low-temperature treatment. The treatment tem-peratures were 4 °C, 0 °C, − 5 °C, − 10 °C and − 15 °C; thenormal environment temperature was used as a control.For temperatures below 0 °C (freezing damage), M. falcataseedlings were acclimated for 2 days at 4 °C and then ex-posed to low-temperature stress. The cold stress inductionschedule involved a decrease of 1 °C every 1 h from 4 °C,and the low-temperature stress treatment was applied ateach studied temperature for 2 h [56, 57]. The roots wereharvested, immediately frozen in liquid nitrogen andstored at − 80 °C for laboratory analysis.

Physiological assays of low-temperature-treated M.falcata rootsRoot cell membrane damage was assessed via electrolyteleakage [58]. The MDA content was measured accordingto the modified thiobarbituric acid (TBA) method [59],and the activity of SOD was measured by the nitro bluetetrazolium (NBT) method [60]. The activity of CATwas measured according to the methods of Maehly andChance [61], and the activity of POD was measured ac-cording to the methods of Zaharieva et al. [62]. The O2

.-

contents were determined as described by Elstner [63],the soluble protein content was determined according tothe Bradford method [64], the content of reduced GSHwas fluorometrically estimated [65], the proline contentwas determined by the ninhydrin method [66], and thesoluble sugar content was determined according to themethods of Dreywood [67].All assays described above were repeated four times,

with four biological replicates. The data, which are shownas the means ± SDs, were subjected to ANOVA to deter-mine significant differences. The least significant differ-ences (LSDs) of the means were determined via Duncan’stest at the level of significance defined as α = 0.05.

RNA isolation, library preparation and sequencingTotal RNA from each sample was isolated via TRIzol re-agent (Invitrogen, USA) according to the manufacturer’s

Cui et al. BMC Plant Biology (2019) 19:575 Page 12 of 16

protocol, and genomic DNA was removed via digestionwith DNase I (TaKaRa, Japan). The purity, concentrationand nucleic acid absorption peak were then measuredwith a Nanodrop ND-1000 spectrophotometer (Nano-Drop, USA). The RNA integrity was measured by anAgilent 2100 Bioanalyzer (Agilent, USA), and genomicDNA contamination was detected by electrophoresis.The library was prepared after the samples passed the

quality tests. For Illumina cDNA library preparation,20 μg of total RNA from each pool was enriched witholigo (dT) magnetic beads and randomly interrupted bythe addition of fragmentation buffer. First-strand cDNAwas then synthesized with random hexamers, withmRNA used as a template. Second-strand cDNA wassynthesized after the addition of buffer, dNTPs, RNaseH and DNA polymerase I. The cDNA was subsequentlypurified with AMPure XP beads. The purified double-stranded cDNA was subjected to end repair, the additionof a poly-A tail and ligation with sequencing linkers, andthe fragment size was selected via AMPure XP beads. Fi-nally, the cDNA library was prepared by PCR-basedenrichment.With respect to PacBio Iso-Seq library preparation, the

cDNA was synthesized using a SMARTer™ PCR cDNASynthesis Kit (TaKaRa, Japan). cDNA libraries of differ-ent sizes were generated by BluePippin. The screenedcDNA was then amplified by PCR, end repaired, con-nected to the SMRT dumbbell-type connector, and exo-nuclease digested. Finally, the library was prepared aftera secondary screening by BluePippin. A total of eightSMRT cells were used for the three libraries at three sizeranges: 1–2 kb, 2–3 kb, and 3–6 kb.After the accurate quantification of libraries via Qubit 2.0

and qualification of the library sizes via an Agilent 2100 in-strument, the libraries were sequenced via PacBio RS II(with 8 SMRT cells) and via the Illumina HiSeq 2500 plat-form at the Biomarker Institute (Biomarker, China). The1–2 kb, 2–3 kb and 3–6 kb libraries were sequenced in con-junction with 3, 3 and 2 SMRT cells, respectively.

Quality filtering and transcriptome assemblyRaw reads were processed into error-corrected ROIs bythe ToFu pipeline, with full passes> = 0, and the accur-acy of the sequence was greater than 0.75 (https://github.com/PacificBiosciences/cDNA_primer/wiki/Un-derstanding-PacBio-transcriptome-data#readexplained).High-quality clean data were obtained by removing readscontaining connectors and low-quality reads (includingthose with an N removal ratio greater than 10% andreads where the number of bases with a mass value ofQ ≤ 10 accounted for more than 50% of the reads).FLNC transcripts were then determined by searching forpoly-A tail signals and the 5′ and 3′ cDNA primers inROIs. Iterative clustering for error correction (ICE) was

used to obtain consensus isoforms, and the full-lengthconsensus sequences from ICE were polished usingSMRT Analysis (version 2.3.0). Full-length transcriptswith a post correction accuracy greater than 99% weregenerated for further analysis. Redundant reads were re-moved from the Iso-Seq™ high-quality full-length tran-scripts by CD-HIT (identity > 0.99). The resultingtranscript sequence was directly used for subsequentanalyses of AS events, SSRs and lncRNAs. The second-generation data were used to quantify and differentiallyanalyze the new CDS.

Identification of AS events, SSRs, lncRNAs and CDSsThe Iso-Seq™ data was used to perform all-vs-allBLAST, and the BLAST alignments met all criteriawere considered as the products of candidate ASevents. The AS gap was larger than 100 bp and atleast 100 bp away from the 3′/5′ end. SSRs withinthe transcriptome were identified by MISA (http://pgrc.ipk-gatersleben.de/misa/). Transcripts withlengths greater than 200 nt and with more than twoexons were selected as lncRNA candidates and furtherscreened via CPC/CNCI/CPAT/Pfam, which have thepower to distinguish protein-coding genes from thenoncoding genes. TransDecoder (https://github.com/TransDecoder/TransDecoder/releases) was used toidentify CDS regions within the transcript sequences.

Gene functional annotationTo acquire the most comprehensive annotation infor-mation, all full-length transcripts identified from theSMRT sequencing were aligned with the NCBI Nr,SwissProt, GO, COG, KOG, Pfam, and KEGG data-bases by BLAST software (version 2.2.26) [68]. GOenrichment analysis was implemented by the GOseqR package on the basis of Wallenius noncentralhypergeometric distribution [69]. KOBAS softwarewas subsequently used to test the statistical enrich-ment of DEGs in the KEGG pathways [70].

Quantification of gene expression levelsThe gene expression level of each sample was identifiedby RSEM [71]. The clean data were mapped back ontothe assembled transcriptome, and read count for eachgene was obtained from the mapping results. Bowtie2software was used to compare the clean data from Illu-mina sequencing to the SMRT sequencing data. Quanti-fication of gene expression levels was estimated viaFPKM values, considering the effects of the sequencingdepth and gene length on the fragments.

Identification of DEGsDifferential expression analysis was performed by theDESeq R package (version 1.10.1) to identify DEGs

Cui et al. BMC Plant Biology (2019) 19:575 Page 13 of 16

between the low-temperature-treated samples and thecontrol samples collected at different temperaturepoints [72]. DESeq performs statistical analyses fordetermining differential expression among digital geneexpression data via a model based on a negative bino-mial distribution. For the detection of DEGs, a foldchange≥2 and an FDR < 0.01 were used as screeningcriteria. The fold change represents the ratio of ex-pression between two samples (groups). The FDR wasobtained by correcting the p values of different sig-nificance. Because the differences in the transcriptomesequencing expression analysis are the transcribed ex-pression values of a large number of independentstatistical hypothesis tests, false positives are a con-cern; thus, in the process of analyzing DEGs, therecognized Benjamini-Hochberg correction method ofhypothesis testing with the original significant pvalues for correction and, subsequently, the FDR wereused as key indicators for screening DEGs.

Identification of putative TFsBLAST was used to search all the DEGs against plant aTF database (Plant TFDB version 4.0, http://planttfdb.cbi.pku.edu.cn) to identify putative TFs. The TF infor-mation was annotated on the basis of the comparisonresults.

Coexpression network analysis with WGCNACoexpression networks were constructed via theWGCNA package in R from all the DEGs [53]. Moduleswere obtained by the automatic network constructionfunction blockwise modules, with the default settings.The eigengene value was calculated for each module andused to test the association with each physiologicalindex. The total connectivity and intramodular connect-ivity (function soft connectivity), kME (for modularmembership), and kME p values were calculated for theDEGs.

Validation of RNA-seq data by qRT-PCRThe RNA samples isolated above were used as templatesand were reverse transcribed with a HiScript II Q SelectRT SuperMix for qPCR (gDNA eraser) kit (Vazyme,China). Primers used in this study were designed via Pri-mer 5 with RefSeq and are listed in Additional file 16:Table S11. The expression of the beta-actin gene wasused as an internal control. qRT-PCR was performedwith ChamQ™ Universal SYBR qPCR Master Mix(Vazyme, China) on a LightCycler 480 II58 device(Roche, Switzerland) according to the manufacturers’protocol. Relative gene expression levels were evaluatedaccording to the 2-ΔΔCT method [73].

Supplementary informationSupplementary information accompanies this paper at https://doi.org/10.1186/s12870-019-2192-1.

Additional file 1: Table S1. Overview of the quality of the sequencedata obtained by NGS sequencing.

Additional file 2: Table S2. Statistics of AS. The QueryName andSubjectName are the IDs of the identified AS events. QhspStart1,QhspEnd1, QhspStart2, QhspEnd2, ShspStart, ShspEnd1, ShspStart2 andShspEnd2 indicate the start and end positions of the 2 HSPs for thesetwo AS events.

Additional file 3: Table S3. Statistics of SSRs.

Additional file 4: Fig. S1. Venn diagram of lncRNAs via CPC, CNCI,CPAT and information in the Pfam database.

Additional file 5: Table S4. Statistics of annotated transcripts.

Additional file 6: Fig. S2. Nr homologous species distribution.

Additional file 7: Table S5. Summary of reads from the RNA-seq dataand their matches with full-length transcripts.

Additional file 8: Table S6. FPKM values and functional annotations ofall transcripts.

Additional file 9: Table S7. Information on the 11,369 genesdifferentially expressed under low-temperature stress.

Additional file 10: Table S8. Summary of the 1538 TFs in response tolow-temperature stress.

Additional file 11: Fig. S3. Expression pattern of TFs. The heat mapshows the log10(FPKM+ 1) values of 1538 individual TFs.

Additional file 12: Table S9. Summary of coexpressed genes betweenlow-temperature-treated samples and control samples.

Additional file 13: Table S10. Genetic modules and kME values ofgenes identified in the WGCNA results.

Additional file 14: Fig. S4. Functional analysis of genes in theMEbrown module. A, COG functional classification. B, GO classification. C,KEGG enrichment.

Additional file 15: Fig. S5. qRT-PCR assay results of genes selected forRNA-seq data confirmation. The blue line represents the normalizedexpression (log10(FPKM+ 1)) of RNA-seq data shown on the Y-axis on theleft. The red line represents the relative qRT-PCR expression level datashown on the Y-axis on the right.

Additional file 16: Table S11. List of primers used in this paper.

AbbreviationsAS: Alternative splicing; CAT: catalase; CDS: coding DNA sequence;CNCI: Coding–non-coding index; COG: Clusters of Orthologous Groups;COLD1: Chilling-tolerance divergence 1; COR: Cold-responsive; CPAT: Codingpotential assessment tool; CPC: Coding potential calculator;DEG: Differentially expressed gene; FDR: False discovery rate; FLNC: full-length nonchimeric read; FPKM: Fragments per kilobase of transcriptsequence per million base pairs sequenced; GO: Gene Ontology;GSH: Reduced glutathione; ICE1: inducer of CBF expression 1; KEGG: KyotoEncyclopedia of Genes and Genomes; KOG: Eukaryotic Ortholog Groups;LEA: Late embryogenesis abundant proteins; lncRNA: Long noncoding RNA;LSD: least significant difference; MDA: Malondialdehyde; NGS: Next-generation sequencing; Nr: NCBI nonredundant protein; O2

.-: Superoxideanion radical; Pfam: Protein family; POD: Peroxidase; qRT-PCR: Quantificational real-time polymerase chain reaction; RNA-seq: RNAsequencing; ROI: Read of inserts; ROS: Reactive oxygen species; RSEM: RNAsequencing by expectation maximization; SMRT: Single-molecule real-time;SOD: Superoxide dismutase; SPS: Sucrose phosphate synthase; SSR: Simplesequence repeat; TF: Transcription factor; WGCNA: Weighted genecoexpression network analysis

AcknowledgmentsWe thank the Biomarker Corporation (Beijing, China) for the SMRTsequencing and RNA-seq and for the raw data analysis.

Cui et al. BMC Plant Biology (2019) 19:575 Page 14 of 16

Authors’ contributionsGC, HC and PZ designed the research. PZ and HC wrote the paper, withcontributions from and discussion with all of the coauthors. HC, HY, MY, GH,MG and RY conducted the research. All authors have read and approved themanuscript.

FundingThis work was financially supported by the National Key R&D Program ofChina (2016YFC0500607), the National Science Foundation of China(31502007), the Young Talents Project of Northeast Agricultural University(16QC23), the Special Funds for Science and Technology Innovation TalentResearch in Harbin (2017RAQXJ008), and the Project of Nature ScientificFoundation of Heilongjiang Province, China (QC2015019). Each of thefunding bodies granted the funds on the basis of a research proposal. Thebodies had no influence on the experimental design, data analysis andinterpretation, or writing of the manuscript.

Availability of data and materialsAll relevant supplementary data are provided within this manuscript asAdditional files. The PacBio SMRT reads and the Illumina NGS readsgenerated in this study were submitted to the NCBI Sequence Read Archiveunder accession numbers BioProject PRJNA 549099 and 520970, respectively.The address is as follows: http://www.ncbi.nlm.nih.gov/sra.

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests.

Author details1Department of Grassland Science, College of Animal Science andTechnology, Northeast Agricultural University, Harbin 150030, China. 2Branchof Animal Husbandry and Veterinary of Heilongjiang Academy of AgriculturalSciences, Qiqihar 161005, China. 3Hulunbuir Grassland Station, Hulunbuir021008, China.

Received: 31 July 2019 Accepted: 8 December 2019

References1. Jin J, Kim J. Cold stress signaling networks in Arabidopsis. J Plant Biol. 2013;

56(2):69–76.2. Zhou MQ, Shen C, Wu LH, Tang KX, Lin J. CBF-dependent signaling

pathway: a key responder to low temperature stress in plants. Crit RevBiotechnol. 2011;31(2):186–92.

3. Wong CE, Li Y, Labbe A, Guevara D, Nuin P, Whitty B, Diaz C, Golding GB,Gray GR, Weretilnyk EA, et al. Transcriptional profiling implicates novelinteractions between abiotic stress and hormonal responses in Thellungiella,a close relative of Arabidopsis. Plant Physiol. 2006;140(4):1437–50.

4. Bg DLR, Yun SJ, Herath V, Xu F, Park MR, Lee JI, Kim KY. Phenotypic,physiological, and molecular evaluation of rice chilling stress response atthe vegetative stage. Methods Mol Biol. 2013;956:227–41.

5. Gehan MA, Greenham K, Mockler TC, Mcclung CR. Transcriptional networks— crops, clocks, and abiotic stress. Curr Opin Plant Biol. 2015;24:39–46.

6. Janska A, Marsik P, Zelenkova S, Ovesna J. Cold stress and acclimation -what is important for metabolic adjustment? Plant Biol (Stuttg). 2010;12(3):395–405.

7. Majlath I, Darko E, Palla B, Nagy Z, Janda T, Szalai G. Reduced light and moderatewater deficiency sustain nitrogen assimilation and sucrose degradation at lowtemperature in durum wheat. J Plant Physiol. 2016;191:149–58.

8. Hu Z, Fan J, Xie Y, Amombo E, Liu A, Gitau MM, Khaldun ABM, Chen L, Fu J.Comparative photosynthetic and metabolic analyses reveal mechanism ofimproved cold stress tolerance in bermudagrass by exogenous melatonin.Plant Physiol Biochem. 2016;100:94–104.

9. Zheng YL, Li WQ, Sun WB. Effects of acclimation and pretreatment withabscisic acid or salicylic acid on tolerance of Trigonobalanus doichangensisto extreme temperatures. Biol Plant. 2015;59(2):382–8.

10. Theocharis A, Clement C, Barka EA. Physiological and molecular changes inplants grown at low temperatures. Planta. 2012;235(6):1091–105.

11. Trischuk RG, Schilling BS, Low NH, Gray GR, Gusta LV. Cold acclimation, de-acclimation and re-acclimation of spring canola, winter canola and winterwheat: the role of carbohydrates, cold-induced stress proteins andvernalization. Environ Exp Bot. 2014;106(1):156–63.

12. Li XD, Zhuang KY, Liu ZM, Yang DY, Ma NN, Meng QW. Overexpression of anovel NAC-type tomato transcription factor, SlNAM1, enhances the chillingstress tolerance of transgenic tobacco. J Plant Physiol. 2016;204:54–65.

13. Zhu JK. Abiotic stress signaling and responses in plants. Cell. 2016;167(2):313–24.14. Shi Y, Ding Y, Yang S. Cold signal transduction and its interplay with

phytohormones during cold acclimation. Plant Cell Physiol. 2015;56(1):7–15.15. Ma Y, Dai X, Xu Y, Luo W, Zheng X, Zeng D, Pan Y, Lin X, Liu H, Zhang D,

et al. COLD1 confers chilling tolerance in rice. Cell. 2015;160(6):1209–21.16. Chinnusamy V, Zhu J, Zhu J-K. Cold stress regulation of gene expression in

plants. Trends Plant Sci. 2007;12(10):444–51.17. Jagloottosen KR, Gilmour SJ, Zarka DG, Schabenberger O, Thomashow MF.

Arabidopsis CBF1 overexpression induces COR genes and enhances freezingtolerance. Science. 1998;280(5360):104–6.

18. Jeknic Z, Pillman KA, Dhillon T, Skinner JS, Veisz O, Cuesta-Marcos A, HayesPM, Jacobs AK, Chen THH, Stockinger EJ. Hv-CBF2A overexpression in barleyaccelerates COR gene transcript accumulation and acquisition of freezingtolerance during cold acclimation. Plant Mol Biol. 2014;84(1–2):67–82.

19. Vanburen R, Bryant D, Edger PP, Tang H, Burgess D, Challabathula D, SpittleK, Hall R, Gu J, Lyons E. Single-molecule sequencing of the desiccation-tolerant grass Oropetium thomaeum. Nature. 2015;527(7579):508.

20. Lan T, Renner T, Ibarra-Laclette E, Farr KM, Chang TH, Cervantes-Pérez SA,Zheng C, Sankoff D, Tang H, Purbojati RW. Long-read sequencing uncoversthe adaptive topography of a carnivorous plant genome. Proc Natl Acad SciU S A. 2017;114(22):4433–41.

21. Chaisson MJ, Huddleston J, Dennis MY, Sudmant PH, Malig M, HormozdiariF, Antonacci F, Surti U, Sandstrom R, Boitano M. Resolving the complexity ofthe human genome using single-molecule sequencing. Nature. 2014;517(7536):608–11.

22. Chen X, Bracht JR, Goldman AD, Dolzhenko E, Clay DM, Swart EC, PerlmanDH, Doak TG, Stuart A, Amemiya CT. The architecture of a scrambledgenome reveals massive levels of genomic rearrangement duringdevelopment. Cell. 2014;158(5):1187–98.

23. Zhang T, Zhao X, Wang W, Pan Y, Huang L, Liu X, Zong Y, Zhu L, Yang D,Fu B. Comparative transcriptome profiling of chilling stress responsivenessin two contrasting rice genotypes. PLoS One. 2012;7(8):e43274.

24. Karki A, Horvath DP, Sutton F. Induction of DREB2A pathway with repression of E2F,jasmonic acid biosynthetic and photosynthesis pathways in cold acclimation-specific freeze-resistant wheat crown. Funct Integr Genomics. 2013;13(1):57–65.

25. Xu W, Li R, Zhang N, Ma F, Jiao Y, Wang Z. Transcriptome profiling of Vitisamurensis, an extremely cold-tolerant Chinese wild Vitis species, revealscandidate genes and events that potentially connected to cold stress. PlantMol Biol. 2014;86(4–5):527–41.

26. Xu Z, Peters RJ, Weirather J, Luo H, Liao B, Zhang X, Zhu Y, Ji A, Zhang B, Hu S,et al. Full-length transcriptome sequences and splice variants obtained by acombination of sequencing platforms applied to different root tissues of Salviamiltiorrhiza and tanshinone biosynthesis. Plant J. 2015;82(6):951–61.

27. Yang LF, Jin YH, Huang W, Sun Q, Liu F, Huang XZ. Full-lengthtranscriptome sequences of ephemeral plant Arabidopsis pumila providesinsight into gene expression dynamics during continuous salt stress. BMCGenomics. 2018;19:717.

28. Chao Y, Yuan J, Li S, Jia S, Han L, Xu L. Analysis of transcripts and spliceisoforms in red clover (Trifolium pratense L.) by single-molecule long-readsequencing. BMC Plant Biol. 2018;18(1):300.

29. Yang Y, Yu Q, Yang Y, Su Y, Ahmad W, Guo J, Gao S, Xu L, Que Y.Identification of cold-related miRNAs in sugarcane by small RNAsequencing and functional analysis of a cold inducible ScmiR393 to coldstress. Environ Exp Bot. 2018;155:464–76.

30. Shen C, Li D, He R, Fang Z, Xia Y, Gao J, Shen H, Cao M. Comparativetranscriptome analysis of RNA-seq data for cold-tolerant and cold-sensitiverice genotypes under cold stress. J Plant Biol. 2014;57(6):337–48.

31. Zhang LL, Zhao MG, Tian QY, Zhang WH. Comparative studies on toleranceof Medicago truncatula and Medicago falcata to freezing. Planta. 2011;234(3):445–57.

32. Riday H, Brummer EC. Forage yield Heterosis in alfalfa. Crop Sci. 2002;42(3):716–23.

Cui et al. BMC Plant Biology (2019) 19:575 Page 15 of 16

33. Gréard C, Barre P, Flajoulot S, Santoni S, Julier B. Sequence diversity of fiveMedicago sativa genes involved in agronomic traits to set up allele miningin breeding. Mol Breed. 2018;38(12):41.

34. Tan J, Zhuo C, Guo Z. Nitric oxide mediates cold- and dehydration-inducedexpression of a novel MfHyPRP that confers tolerance to abiotic stress.Physiol Plant. 2013;149(3):310–20.

35. Qu YT, Duan M, Zhang ZQ, Dong JL, Wang T. Overexpression of theMedicago falcata NAC transcription factor MfNAC3 enhances cold tolerancein Medicago truncatula. Environ Exp Bot. 2016;129:67–76.

36. Miao Z, Xu W, Li D, Hu X, Liu J, Zhang R, Tong Z, Dong J, Su Z, Zhang L,et al. De novo transcriptome analysis of Medicago falcata reveals novelinsights about the mechanisms underlying abiotic stress-responsivepathway. BMC Genomics. 2015;16:818.

37. Liu M, Wang TZ, Zhang WH. Sodium extrusion associated with enhancedexpression of SOS1 underlies different salt tolerance between Medicagofalcata and Medicago truncatula seedlings. Environ Exp Bot. 2015;110:46–55.

38. Chen H, Chen X, Chen D, Li J, Zhang Y, Wang A. A comparison of the lowtemperature transcriptomes of two tomato genotypes that differ in freezingtolerance: Solanum lycopersicum and Solanum habrochaites. BMC Plant Biol.2015;15(1):132.

39. Tondelli A, Francia E, Barabaschi D, Pasquariello M, Pecchioni N. Inside theCBF locus in Poaceae. Plant Sci. 2011;180(1):39–45.

40. Dubé M-P, Castonguay Y, Cloutier J, Michaud J, Bertrand A. Characterizationof two novel cold-inducible K3 dehydrin genes from alfalfa (Medicago sativaspp. sativa L.). Theor Appl Genet. 2013;126(3):823–35.

41. Jithesh MN, Prashanth SR, Sivaprakash KR, Parida AK. Antioxidative responsemechanisms in halophytes: their role in stress defence. J Genet. 2006;85(3):237–54.

42. Mittler R, Vanderauwera S, Gollery M, Van BF. Reactive oxygen gene networkof plants. Trends Plant Sci. 2004;9(10):490–8.

43. Huot B, Castroverde CDM, Velasquez AC, Hubbard E, Pulman JA, Yao J,Childs KL, Tsuda K, Montgomery BL, He SY. Dual impact of elevatedtemperature on plant defence and bacterial virulence in Arabidopsis. NatCommun. 2017;8(1):1808.

44. Mohapatra SS, Wolfraim L, Poole RJ, Dhindsa RS. Molecular cloning andrelationship to freezing tolerance of cold-acclimation-specific genes ofalfalfa. Plant Physiol. 1989;89(1):375–80.

45. Fan L, Li R, Pan J, Ding Z, Lin J. Endocytosis and its regulation in plants.Trends Plant Sci. 2015;20(6):388–97.

46. Leborgne-Castel N, Luu DT. Regulation of endocytosis by external stimuli inplant cells. Plant Biosystems. 2009;143(3):630–5.

47. Rihan HZ, Al-Issawi M, Fuller MP. Advances in physiological and molecularaspects of plant cold tolerance. J Plant Interact. 2017;12(1):143–57.

48. Pan H, Wang Y, Zhang Y, Zhou T, Fang C, Nan P, Wang X, Li X, Wei Y, Chen J.Phenylalanine ammonia lyase functions as a switch directly controlling theaccumulation of calycosin and calycosin-7-O-beta-D-glucoside in Astragalusmembranaceus var. mongholicus plants. J Exp Bot. 2008;59(11):3027–37.

49. Hummel I, Bourdais G, Gouesbet G, Couee I, Malmberg RL, El Amrani A.Differential gene expression of ARGININE DECARBOXYLASE ADC1 and ADC2 inArabidopsis thaliana: characterization of transcriptional regulation during seedgermination and seedling development. New Phytol. 2004;163(3):519–31.

50. Liu X, Bush DR. Expression and transcriptional regulation of amino acidtransporters in plants. Amino Acids. 2006;30(2):113–20.

51. Zhu XC, Song FB, Liu FL. Altered amino acid profile of arbuscularmycorrhizal maize plants under low temperature stress. J Plant Nutr Soil Sci.2016;179(2):186–9.

52. Saxena SC, Kaur H, Verma P, Petla BP, Majee M. Osmoprotectants: potentialfor crop improvement under adverse conditions. 2013.

53. Langfelder P, Horvath S. WGCNA: an R package for weighted correlationnetwork analysis. BMC Bioinformatics. 2008;9(1):559.

54. Zhao W, Langfelder P, Fuller T, Dong J, Li A, Hovarth S. Weighted genecoexpression network analysis: state of the art. J Biopharm Stat. 2010;20(2):281–300.

55. Hoagland D, Arnon D. The water culture method for growing plantswithout soil. Calif Agr Expt S. 1950;347(1):32.

56. Iii DPL, Tan DTJE, Botany E. Understanding the response of winter cereals tofreezing stress through freeze-fixation and 3D reconstruction of iceformation in crowns. Environ Exp Bot. 2014;106(1):24–33.

57. Verslues PE, Agarwal M, Katiyar-Agarwal S, Zhu J, Zhu JK. Methods andconcepts in quantifying resistance to drought, salt and freezing, abioticstresses that affect plant water status. Plant J. 2006;45(4):523–39.

58. Peever TL, Higgins VJ. Electrolyte leakage, Lipoxygenase, and lipidperoxidation induced in tomato leaf tissue by specific and nonspecificelicitors from Cladosporium fulvum. Plant Physiol. 1989;90(3):867–75.

59. Wang WB, Kim YH, Lee HS, Kim KY, Deng XP, Kwak SS. Analysis ofantioxidant enzyme activity during germination of alfalfa under salt anddrought stresses. Plant Physiol Biochem. 2009;47(7):570–7.

60. Giannopolitis CN, Ries SK. Superoxide dismutases: I. occurrence in higherplants. Plant Physiol. 1977;59(2):309–14.

61. Maehly AC. The assay of catalases and peroxidases. In: Glick D, editor.Methods biochemical analysis. Hoboken: Wiley; 2006. p. 357–424.

62. Zaharieva T, Yamashita K, Matsumoto H. Iron deficiency induced changes inAscorbate content and enzyme activities related to Ascorbate metabolismin cucumber roots. Plant Cell Physiol. 1999;40(3):273–80.

63. Elstner EF, Heupel A. Inhibition of nitrite formation fromhydroxylammoniumchloride: a simple assay for superoxide dismutase. AnalBiochem. 1976;70(2):616–20.

64. Bradford MM. A rapid and sensitive method for the quantitation ofmicrogram quantities of protein utilizing the principle of protein-dyebinding. Anal Biochem. 1976;72(1–2):248–54.

65. Hissin PJ, Hilf R. A fluorometric method for determination of oxidized andreduced glutathione in tissues. Anal Biochem. 1976;74(1):214–26.

66. Bates LS, Waldren RP, Teare ID. Rapid determination of free proline forwater-stress studies. Plant Soil. 1973;39(1):205–7.

67. Dreywood R. Qualitative test for carbohydrate material. Ind Eng Chem AnalEd. 1946;18(8):499.

68. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ.Gapped BLAST and PSI-BLAST: a new generation of protein database searchprograms. Nucleic Acids Res. 1997;25(17):3389–402.

69. Young MD, Wakefield MJ, Smyth GK, Oshlack A. Gene ontology analysis forRNA-seq: accounting for selection bias. Genome Biol. 2010;11(2):R14.

70. Xie C, Mao X, Huang J, Ding Y, Wu J, Dong S, Kong L, Gao G, Li CY, Wei L.KOBAS 2.0: a web server for annotation and identification of enrichedpathways and diseases. Nucleic Acids Res. 2011;39(Web Server issue):W316–22.

71. Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq datawith or without a reference genome. BMC Bioinformatics. 2011;12:323.

72. Anders S, Huber W. Differential expression analysis for sequence count data.Genome Biol. 2010;11(10):R106.

73. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2 ΔΔC T method. Methods. 2001;25:402–8.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Cui et al. BMC Plant Biology (2019) 19:575 Page 16 of 16